Nvidia takes on Intel, ventures into business apps
Nvidia is attempting to break out of the mould of media and technical processing by developing graphics chips geared towards business applications.
Maxwell is the next big step on Nvidia’s road map, a graphics chip due in 2014 using virtual memory shared by the GPU and host CPU.
The approach, likely implemented in Nvidia’s proprietary Cuda environment, is similar to what Advanced Micro Devices is enabling with its HSA group supported by ARM, Microsoft and others.
In 2015, Nvidia will pack a Maxwell core along with its first 64bit ARM design into a sixth-generation Tegra chip called Parker.
The device is the first of Nvidia’s Project Denver designs and will be made in a FinFET process, said Jen-Hsun Huang, Nvidia’s chief executive in a keynote here.
Separately, Nvidia will sample Tegra5, aka Logan, this year, its first mobile processor to support its Cuda environment for general-purpose graphics. It also supports OpenGL 4.3 and will be in production in 2014.
The company also disclosed a 2015 graphics processor called Volta, its first to use stacked memory with through-silicon vias, similar to the Hyper Memory Cube defined by Micron and supported by a host of other chip makers. Also at GTC, Nvidia is rolling out its Grid Enterprise initiative. It consists of a new layer of systems and software to provide business users virtualized access to apps running on Nvidia graphics processors.
Dell, Hewlett-Packard and IBM are among companies who will make Grid Enterprise servers and appliances for large and small businesses. Citrix, Microsoft and VMWare are among those making the software.
The move is a small step towards greater head-to-head competition between the graphics chip maker and the x86 giant. In the background, Nvidia is still developing its Project Denver a suit of ARM-based SoCs intended to enable a full range of systems from notebooks to supercomputers.
Nvidia will pave the way for such chips this week, announcing its Cuda software for general-purpose graphics (GPGPUs) computing is now running on ARM processors. At least one supercomputer developer will talk about using Cuda on ARM at GTC.
Dozens of apps for GPGPUs
Top supercomputers have been adopting GPGPUs for years, and ARM server SoCs are starting to gain limited traction. It’s not clear exactly when the first Project Denver chips will ship, attacking more mainstream computing markets.
Meanwhile, Nvidia’s GTC will showcase dozens of apps for GPGPUs from iRobot vacuum cleaners to cars from Audi, BMW and Chrysler who will have executives speaking at the event.
Among other new business uses, Salesforce.com is using Nvidia chips to analyse Twitter feeds and Shazam uses the chips to run online searches for music. Cortexica Visual Systems has an app that lets consumers take pictures of clothes and find similar items for sale online.
The base of programmers enabled to use Nvidia’s graphics programming environment is expanding. Continuum Analytics announced NumbaPro, a Python compiler for Cuda in the new Anaconda Accelerate suite.
Nvidia claims the software opens the door to as many as 50,000 new Python programmers trained and able to use Cuda within five months, thanks to new online courses.
Observers say Nvidia’s Cuda currently has the lead in the emerging GPGPU field due to its maturity. Nvidia early on seeded universities with classes on Cuda.
Long term, many expect the open source OpenCL environment will dominate the GPGPU field. It is supported by all graphics players including AMD and Intel as well as Nvidia, but it not yet nearl